TRIBOLOGY IN RELIABILITY ENGINEERING

K. HOLMBERG
VTT Manufacturing Technology, P. O. Box 1704, FIN-02044 VTT, FINLAND; e-mail: kenneth.holmberg@vtt.fi

SUMMARY
The tribological understanding of friction and wear mechanisms and the generation of reliable friction and wear data are
of increasing importance for the determination of reliability of machinery and availability of plants and transportation
systems. To be of real value in reliability engineering, the tribological data should be expressed in terms of endurance
life and probability of failure. The minimum coefficient of friction and the minimum wear data are important
scientifically and for optimal design and material development purposes, but they are of less relevance in reliability
estimation. The scaling-up of our tribological knowledge from nano- and microcontacts to solid performance estimations
and reliability predictions of large machines is a major challenge for the tribological society.

Keywords: reliability, tribology, design, monitoring, scaling

1 INTRODUCTION engineering, the different tribology-related methods that

exists to improve reliability, and finally the scope of
A general trend in our society is to develop larger and
tribology covering technical and scientific problems
more advanced technical systems which are at the same
from nanoscale up to terascale.
time more complex and difficult to control. The
availability of plants and transportation systems and the 2 A HOLISTIC APPROACH TO
reliability of machines and instruments is of increasing RELIABILITY
importance. The failure of technical systems may result
Since there is a great variety of different techniques
in safety risks, and in large environmental risks.
based on expert knowledge in the several fields of
Breakdowns in the industrial production process are
technology involved, there is a need to approach the
expensive for the industry and the whole society. There
reliability and maintainability problems from a general
is an increasing need for improved methods of
holistic point of view starting from the problem of the
determining the reliability and predicting the lifetime of
customer and ending with the satisfied user [4, 6]. This
machines and production systems more accurately.
is aimed at improving the synergistic interactions
Mechanical failures of components and especially between the different fields of expertise by showing a
tribological failures, such as wear and friction-related logical and comprehensive structure, where each expert
failures, are today one of the main reasons for can find his place and see the connections to experts
machinery break-down, plant shut-down and equipment from other fields, all working with the same aim of well
unavailability [1−5]. This paper deals with the role of controlled risks, failure frequency and lifetime, as shown
tribology in the large and complex scope of reliability in Figure 1.

RISK CONTROL CORRECTIVE ACTION

ANALYSIS IMPROVEMENT - change of component
- improved design
Probability of personal-, equipment- - monitoring
and environmental damage - automatic diagnostics
Accident consequence estimation - inspections
- service
- redundancy
RELIABILITY CONTROL
- operational tests
ANALYSIS OPTIMIZATION

Identification of critical parts

System failure- and lifetime probability RISK ESTIMATION
Estimation of operability costs (LCC) FAILURE
PROBABILITY
LIFETIME
HUMAN ERROR CONTROL ESTIMATION
SOFTWARE FAILURE CONTROL
ELECTRONICS FAILURE CONTROL
MECHANICAL COMPONENT FAILURE CONTROL
WEAR CORROSION CREEP FATIGUE FRACTURE

k hfi 00/fi 1 d f
Figure 1: A holistic systematic approach to the improvement of the operational reliability, availability and
safety of products and industrial production systems.
3 COMPONENT RELIABILITY design specification; this situation results in geometrical
changes and changes in tolerances between components,
The basic tribological phenomena –friction, wear and
and thus in possible inaccurate or even hazardous
lubrication – are all reliability-related. Friction in
movements or failures and shut-down.
machinery and equipment resists motion and can result
in energy losses and failure. Wear is a process of What is important from a reliability point of view is to
material deterioration that will result in failure if it can produce more and better tribological data on the
proceed far enough. Lubrication is the method of endurance life and on the critical wear and friction
controlling friction and wear by introducing a third levels of tribological components, both in general and in
material between the interacting moving surfaces, which specific operational conditions.
can reduce energy losses and help avoid failure [5, 7].
Figure 2a illustrates schematically three situations of
One major challenge for the tribology society is: How component wear. In the first situation, severe wear is
can we produce better tribological data of interacting starting already at an early stage of the operation due to
components to support actions for improved reliability? unsuccessful design. In the second situation a mild wear
In this context, it is not so important to find out what is condition is achieved after a running-in wear period and
the minimum coefficient of friction that can be achieved a considerable safe operating time prevails until the
in a certain case, since it is not the situation with slowly proceeding wear damage of the surface results in
minimum friction that is problematic. The risks of severe wear.
failure and shut-down appear when the design value of a Finally, in the third situation successful design has
coefficient of friction is exceeded, thus creating higher produced a contact condition, which, after a short
frictional forces that may cause damage or stop intended running-in wear, results in a long period of insignificant
mechanical movements. Similarly, the low-wear situa- wear. The repeated contact load will ultimately result,
tion is not the problem, but a problem may occur when however, in fatigue failure and catastrophic wear [8].
the wear rate for some reason exceeds the values in the

Endurance life
CUMULATIVE WEAR

Catastrophic wear
ear
re w
e
Sev

Critical wear level Mild wear

Insignificant wear

TIME

Endurance life

Critical friction level

FRICTION

TIME

Figure 2: An increase in wear and friction may with time exceed the critical wear and friction levels and
then the operation is in the risk zone for total failure and shut-down.
The important question from a reliability point of view the number of fluctuating variables can be limited to a
is to produce tribological data and techniques with controllable level.
which the endurance life can be estimated. The
The laboratory testing of material combinations in
endurance life is the time until the component can no
specific contact situations and environments is today
longer perform its planned function. The designed
perhaps the most reliable way of gaining basic
endurance life is the time until the cumulative wear limit
tribological data for reliability control purposes. It is
for safe operation specified in the design specification,
essential that the experiments are planned and
the critical wear level, is exceeded in operation.
performed specifically for estimating endurance life, as
Similarly, as shown in Figure 2b, the changed contact
discussed above. In addition to endurance life, another
conditions can result in exceeding the endurance life
reliability parameter that is important from a practical
when an increase in the friction force exceeds the values
point of view is the probability of failure. The
in the design specification, the critical friction level, and
tribological data can be of crucial value to the machine
a risk for safe operation occurs.
users when both the endurance life and the probability
of failure can be expressed in statistical terms. The
4 TRIBODATA FOR RELIABILITY established experimental design techniques [3, 12] can
ASSESSMENT be put to good use here, and they should be utilised in
tribology laboratories much more.
There are four ways of getting the required tribological
data for the estimation of component endurance life and When more reliability-related tribological data is
risk for failure. They are: by collecting historical data available, it makes it possible to classify it and to
about how similar components have performed in systematically present it in similar forms as have been
similar operational situations earlier; by using the done in the mapping of friction and wear data [13,14].
generic knowledge of tribological behaviour in The maps are useful when we have first managed to
theoretical models and equations; by laboratory testing identify the two most dominant parameters on friction or
of components and materials, and by on-line monitoring wear and then produced the map with these two
of the tribological performance. parameters as co-ordinate axes. Then we can place in
the map the window for planned operational conditions
There is a lot of knowledge and experience to be gained
defined in the design specification. It is obvious that for
from observations of how similar components have
safe operation there should be no serious transition
performed earlier in similar conditions. One problem is
regions between different contact mechanisms in the
of course that often the earlier operational conditions
operational window. Defining the operational window in
may be more or less different from those to be
tribological performance maps makes it possible to
estimated, and thus the interpretation of the results
calculate the expected variations in endurance life and
includes uncertainties. But the main problem is,
the probability of failure within this window, and if
however, that today in almost all areas of industry this
required, also when moving outside the operational
information is not collected in such a systematic way
window.
that it can easily be used for endurance life prediction.
The history data of used components is often collected The aim should be that we can express our tribological
mainly for the use of legal safety or compensation- data in a way such as: "the endurance life of the
related needs. To be useful, this data should be collected component will be more than 8 years with a 95% degree
by using operational and failure codes especially of confidence" and "the probability of operation without
developed for reliability estimation purposes [1, 9]. failure during the first three years is more than 98%".
With this kind of basic reliability data of components it
The quantitative friction and wear modelling that is
would be possible to estimate with a good degree of
possible to do based on today's state-of-the-art
accuracy the reliability of machines and the availability
techniques has a large variation in the level of accuracy.
of machinery systems and plants. This information is
In some areas, such as in the calculation of surface
essential in situations when decisions on new
stresses or in the calculation of elastohydrodynamic
investments, investments in diagnostic tools and
contact pressures and film thickness, there are quite
redundancy equipment, and decisions on unscheduled
advanced and precise mathematically formulated
shut-downs for repair have to be made.
models. On the other hand, there is no model available
for the estimation of friction or wear in contacts with
5 CONDITION MONITORING AND
coated surfaces in the presence of a fluid containing
contamination particles. Although a great number of
DIAGNOSTICS
equations for friction and wear in dry sliding situations During operation of machinery, actions can be taken to
have been proposed, their field of validity is still limited improve the reliability and tribology has an important
to only some certain contact situations [10,11]. Here, the role to play in this. By condition monitoring of the
most promising approach for producing useful performance, on-line information is obtained about the
tribological data for reliability estimations is, in the stage of functional deterioration that may be due to, e.g.
authors' opinion, not to use generic wear equations but the wear of surfaces. The area of condition monitoring
to develop component-specific friction and wear and diagnostic engineering today is an advanced and
models. In such a way, the geometrical and material- highly developed field of engineering [1, 2]. It includes
related parameters can be fixed to a certain degree, and the development of sensors to measure changes in
performance, data collection, signal processing, tribological understanding about the mechanisms behind
diagnostics, prognostics and maintenance engineering, the functional deterioration observed can reliable
as shown in Figure 3. predictions be made.
Particularly in the area of diagnostics, many new The most common machinery condition monitoring
software tools have been developed for automatically techniques used today are vibration analysis and oil
performing appropriate conclusions based on the analysis. These are both very much related to the
received on-line performance information [15]. Such tribological contact phenomena and a good tribological
new tools are expert systems, fuzzy logic, pattern understanding is essential. Other condition monitoring
recognition and neural networks. The next step after techniques are performance measurements, such as
having made a proper diagnosis of the condition of the measuring rotational speeds, forces, displacements,
machine is to predict how the performance will proceed temperatures etc. [2].
depending on what measures are to be implemented.
This kind of prognostics is very important support for
decision-making in situations where the consequences 6 RELIABILITY-BASED DESIGN
may be great or serious. The last part of the chain is The design methods in use today are mainly based on
maintenance engineering, which includes the knowledge performance calculations and static or dynamic load and
of different corrective actions that can be taken based on durability analysis. New tools have recently been
the information obtained. developed that also take into account the reliability
However, the whole chain from the signal detection by aspects in design. One such tool is the reliability-based
the sensor to prognostics and maintenance decision- design method developed by Virtanen [9,16]. The
making is of relevance only if the analysis of the signals method enables the effective prediction of the reliability
is based on a solid knowledge of the phenomena of a product in explicit form right at the design stage and
observed. This is where tribology comes in. It is most to use it, e.g. during the contract negotiations with a
important in the case of friction and wear-related customer. This method provides a design team with the
monitoring that a solid understanding of the tribological means to judge new and innovative ideas on the basis of
contact mechanisms forms the basis for the signal their predicted reliability, and it helps the team to focus
processing, the diagnostic rules and the prognostic on the right target with their reliability engineering
predictions. Unfortunately, this is not always the case efforts. The method also offers the possibility to analyse
today. Too often the engineers are satisfied with the customer’s view of reliability requirements and to
managing to monitor some kind of signal and then compare this with what is technically possible to
believing that advanced software techniques can achieve, as shown in Figure 4.
produce correct conclusions. Only by bringing good

STOP START

FAILURE DETECTION DATA DIAGNOSTICS, CONCLUSIONS,

Wear COLLECTION PROGNOSTICS ACTIONS
Sequence Fatigue
Fracture
Corrosion

Material science Sensor Signal Diagnostic Maintenance

Physics technology analysis engineering engineering
Knowledge Mechanical Statistics Decision-
needed engineering Information making
technology

Figure 3: Condition monitoring and diagnostic engineering is a multi-technological task. Tribology is involved
in the failure mechanisms, diagnostic analysis and the prognostic predictions.
The right-hand side of Figure 4 shows the process for components and machines. It is actually not very useful
specifying the reliability requirements and priorities of at all if only the minimum achieved coefficient of
the customer for a specific product. This is quantified in friction and the wear rate are known.
terms of reliability parameters such as mean-time-to-
failure (MTTF), mean-time-to-repair (MTTR), avail- 7 FROM NANOTRIBOLOGY TO
ability and maximum breakdowns. The left-hand side TERATRIBOLOGY
shows the process for calculating the same reliability
Relating tribology to machinery reliability and to plant
parameters for the product based on the technical
or transportation availability is very important and
performance of the component and the system. This is a
responds to the increasing demands of both industry and
model-based simulation approach that starts with
society. However, it is a great challenge to scale up the
building up a fault tree for the product. The possible
knowledge we have on the friction and wear
failures are estimated and classified according to best,
mechanisms from tribological studies, which are
mean and worst cases of probable occurrence.
typically made on a micrometer scale in tribometers and
The fault tree that forms the basis for the reliability microscopes, to component and machinery systems that
simulation is based on failure data, diagnosis and can be as large as ferry boats or paper machines.
prediction models. In the case of wear or friction-related
In tribology today we can see another trend, which is to
failures, the estimation of the probability to appear
go to smaller and smaller sizes in the investigation of
during certain periods in the lifetime of the component
friction and wear phenomena. Nanotribology is of
and the estimation of the probable endurance lifetime is
increasing interest for tribologists and here the scales are
based on the best available tribological data. This data
coming down to the very basic dimensions of physical
can be obtained from historical data analysis, from
elements such as atoms and molecules [17]. Emerging
theoretical models or from laboratory-produced data, as
technologies such as the atomic force microscope and
described above. To be of use for the model, the data
the surface force apparatus [18] have recently opened
should be presented as reliability-related parameters.
the possibility to study friction and wear phenomena on
This emphasizes the importance for the tribology society a molecular scale and to measure frictional forces
to produce friction and wear data for different material between contacting atoms at a nano-newton level.
combinations, as well as in different contact and Increased computational power has made it possible to
environmental conditions in the form of endurance life study friction and associated phenomena by molecular
or probability of failure. The data can then be of use in dynamic simulations of sliding surfaces and to
reliability estimations and functional predictions of investigate atomic-scale contact mechanisms [19].

Devising the reliability model Allocation of dependability requirements

Design engineer's Empirical Customer needs i.e. Building of

view of failure failure reasons dependability functional
reasons requirements hierarchy

Technical
solutions Allocation of
Building of the
requirements to
fault tree
functions
Improvement
and re-design of
Design engineer's product features
view of failure
reasons
RAM-requirements:
Simulation - MTTF
model Simulation of - MTTR
reliability and Conclusions - Availibility
sensitivity analysis
Experience and
generic data
banks Comparison of
Simulation run W32 GD Fuel Injection System Power Unit
simulation
"Mean" "Best" "Worst" MTTF MTTR A Max Breakdowns
MTTF 723456 937223 506543
results and 946275 13.6 0.999894 4
MTTR 12.5 13.6 12.3 requirements 473138 6.8 0.999894 8
Availability 0.99976 0.99988 0.99955

Figure 4: A novel reliability-based design method consisting of an advising reliability model and
an allocation model for the customers dependability requirements [9].
The increasing understanding of the tribological Machinery tribology or unitribology describes the
phenomena on nanoscale creates the need to scale up performance-related phenomena for a system of
our nanoscale knowledge to conclusions on improved components assembled in a machine or a piece of
prediction of friction and wear that takes place on a equipment. The parameters of interest are performance,
more macroscopic scale, such as in practically efficiency, reliability and lifetime estimation.
observable everyday life. An attempt to illustrate the
scaling up of the very basic friction and wear pheno- Plant tribology or kilotribology deals with a whole
mena all from atomic dimensions to global and universal system of machinery, structures and equipment – and
dimensions is shown in Figure 5 [20]. now parameters such as economy, risk levels,
availability and life-cycle costs are considered.
Nanotribology could also be called molecular tribology
because here the investigations concentrate on pheno- National tribology or megatribology extends the effects
mena related to the interaction between molecules and and consequences to be seen from a nationwide
atoms such as the effects of van der Waal´s forces and perspective, with the parameters of relevance such as
single crystal structures of materials. safety policy, research policy, transportation policy and
Microtribology or asperity tribology was introduced by environmental policy.
Bowden and Tabor [21] with their studies of friction, Global tribology or gigatribology considers the effects
wear and adhesion that take place at the peaks of the on a worldwide basis as one interacting system; the
surface topography. Phenomena such as fracture, elastic effects that are dealt with are sustainable development,
and plastic deformation, debris formation, surface layer politics and cultural and human survival.
formation and topography effects are of central
importance. Universe tribology or teratribology is the largest per-
Macrotribology or contact-tribology was in focus in the spective today that the author can think of in this scaling
research at the beginning of the last century. These up exercise. But what does this mean? Is it the inter-
works are related to contacts between gears, bearing relations of material in our large space? Is it the mecha-
elements and rollers, and phenomena like Hertzian nisms for space enlarging, or is it mechanisms for the
contact pressure, elastohydrodynamic lubrication, and creation of new life and cultures?
wear mechanisms clearly observable by the naked eye The scaling up from molecular phenomena all the way
(scuffing, scoring, pitting) are of interest. up to universe tribology is an interesting exercise
Component tribology or decitribology is related to de- because it shows that at all levels we have to do with
fining and measuring typical parameters originating phenomena that are known to us to some extent. In this
from the interaction of components and related to their sense, then, it shows that talking about ‘scaling up’ and
performance such as torque, forces, vibrations, clearance ‘scaling down’ is relevant.
and alignment.

1012 Tm Universe tribology - Teratribology

???

109 Gm Global tribology - Gigatribology

Sustainable development, politics, survival

National tribology - Megatribology

106 Mm Safety, environmental, transportation and research policies

103 Km Plant tribology - Kilotribology

Economy, risk levels, availibility, life-cycle cost

1 m Machinery tribology - Unitribology

Performance, efficiency, reliability, lifetime

10-1 dm Component tribology - Decitribology

Torque, forces, vibrations, clearance, alignment

10-3 mm Contact tribology - Macrotribology

Hertz contact pressure, EHD, pitting, scuffing, scoring

10-6 µm Asperity tribology - Microtribology

Fracture, elastic and plastic deformation, debris, surface layers

Molecular tribology - Nanotribology

10-9 nm Molecules, atoms, van der Waal's forces, single crystals, strings

k hfi 00/fi 8 d f
Figure 5: Scaling up of tribological phenomena from nanotribology to teratribology.
However, this exercise also has a considerable risk. It Monitoring and Diagnostic Engineering Management
may give the impression that our whole universe is just a COMADEM´98, Launceston, Australia, 8−11 Dec.
mechanical system, like a complex mechanical clock, 1998, Monash University, Clayton, Australia.
and if we only knew all the details on the smallest [8] Holmberg, K.: Tribological bases for accelerated
atomic and even subatomic level and knew all the laws testing. In: Operational Reliability and Systematic
of interaction then we could calculate and predict what Maintenance. London: Elsevier Applied Science 1991,
takes place nationwide and on a global scale. This is by 31−50.
no means the intention of the author. On the contrary,
there are today indications and opinions that criticize [9] Holmberg, K. (Ed.): Competitive Reliability 1996-
such a mechanistic approach [20, 22, 23]. 2000. Helsinki, Finland: TEKES, National Technology
Agency of Finland, Technology Programme Report 5/
8 CONCLUSIONS 2001.
[10] Holmberg, K. & Matthews, A.: Coatings Tribology.
To be of real value in reliability engineering, the
Amsterdam: Elsevier 1994. Elsevier Tribology Series 28.
tribological data should be expressed in terms of
endurance life and probability of failure. The minimum [11] Meng, H. C. & Ludema, K. C.: Wear models and
coefficient of friction and the minimum wear data are predictive equations: their form and content. Wear, 181-
important scientifically as well as for the optimal design 183 (1995), 443−457.
and material development purpose, but they are of less [12] Montgomery, D. C.: Design and Analysis of
relevance for reliability estimation. The aim is to Experiments. New York: John Wiley & Sons 1976.
transform the tribodata to reliability parameters such as [13] Lim, S. C. & Ashby, M. F.: Wear mechanism maps.
mean-time-to-failure (MTTF), mean-time-to-repair Acta Metall., 35 (1987) 1, 1−24.
(MTTR), availability and maximum number of break-
downs per time. [14] Adachi, K., Kato, K. & Chen, N.: Wear maps of
ceramics. Wear, 203-204 (1997), 291-301.
It is possible to express tribological effects on different
[15] Pusey, H. & Rao, R. (Eds): COMADEM 2000.
geometrical scales all the way from nanotribology,
Proc. 13th International Congress on Condition
perhaps even up to teratribology. However the foreseen
Monitoring and Diagnostic Engineering Management,
scaling-up of our knowledge will probably need other
Houston, USA, 3−8 Dec. 2000. Soc. for Machinery
more nonlinear approaches than the strict reductionism
Failure Prediction Technology, Haymarket, Virginia,
to which we are accustomed.
USA.
9 REFERENCES [16] Virtanen, S.: Reliability in product design −
Specification of dependability requirements. Int.
[1] Holmberg, K. & Folkesson, A. (Eds.): Operational
Symposium on Product Quality and Integrity − The
Reliability and Systematic Maintenance. London:
Reliability and Maintainability Symposium (RAMS).
Elsevier Applied Science 1991.
Anaheim, USA, 19−22 January 1998.
[2] Rao, B. K. N.: Handbook of Condition Monitoring.
Oxford: Elsevier Advanced Technology 1996. [17] Krim, J.: Friction at the atomic scale. Scientific
American, (1996) October, 48−56.
[3] Ireson, W. G., Coombs, C. F. & Moss, R. Y. (Eds.):
Handbook of Reliability Engineering and Maintenance. [18] Israelachvili, J. N. & Tabor, D.: The measurement
New York: McGraw-Hill 1996. of van der Waal´s dispersion forces in the range of 1.5
to 130 nm. Proc. Royal Society, A331 (1972), 19-38.
[4] Holmberg, K.: Advanced solutions for operational
reliability improvements. Proc. 10th Int. Congress on [19] Harrison, J. A., White, C. T., Colton, R. J. &
Condition Monitoring and Diagnostic Engineering Brenner, D. W.: Molecular-dynamics simulations of
Management COMADEM´97, Espoo, Finland, 9−11 atomic-scale friction of diamond surfaces. The
June 1997. Espoo: VTT Technical Research Centre of American Physical Society, Physical Letters B, 46,
Finland. VTT Symposium 171. (1992) 15, 9700-9708.
[5] Kimura, Y.: Tribology as a maintenance tool. In: [20] Holmberg, K.: Reliability aspects on tribology.
New Directions in Tribology. London: Mechanical Tribology International. Accepted for publication 2001.
Engineering Publ. Ltd. 1997, 299−308. 14 p.
[6] Holmberg, K., Vuori, M., Toola, A. & Säynätjoki, [21] Bowden, F. P. & Tabor, D.: Friction and lubrication
M.: A hyperbook supported holistic approach to prob- of solids. Part. I. Oxford: Oxford U. 1950.
lem solving in maintenance and reliability engineering. [22] Goldenfeld, N. & Kadanoff, L. P.: Simple lessons
Proc. 8th Int. Congress Condition Monitoring and from complexity. Science, 2.4.1999, 284, 87-89.
Diagnostic Engineering Management COMADEM´95, [23] Winer, W. O. & Gwaltney, E. C.: Trends and
Kingston, Canada, 26−28 June 1995, Vol. 2. Queen´s directions for tribology research and development in the
University, Kingston, Canada. coming years. Proc. Int. Colloquium Tribology 2000-
[7] Holmberg, K.: Tribology in reliability design and Plus, Esslingen, Germany, 11−13 January 2000,
operational control. Proc. 11th Int. Congress Condition Technische Akademie Esslingen, Esslingen, Germany.